Premium
Robust Predictive Design of Field Measurements for Evapotranspiration Barriers Using Universal Multiple linear Regression
Author(s) -
Clutter Melissa,
Ferré Ty P. A.,
Zhang Zhuanfang Fred,
Gupta Hoshin
Publication year - 2019
Publication title -
water resources research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.863
H-Index - 217
eISSN - 1944-7973
pISSN - 0043-1397
DOI - 10.1029/2019wr026194
Subject(s) - data mining , computer science , field (mathematics) , linear regression , sampling (signal processing) , evapotranspiration , data set , regression , set (abstract data type) , environmental science , machine learning , statistics , artificial intelligence , mathematics , filter (signal processing) , pure mathematics , computer vision , biology , programming language , ecology
Surface barriers are commonly installed to reduce downward water movement into contaminated zones. Specifically, evapotranspiration (ET) barriers are used to store water and release it, via ET, before it can percolate into an underlying waste zone. To assess the effectiveness of a surface barrier, we used an existing data set, model‐simulated data, and a dimensionality reduction approach called universal multiple linear regression (uMLR) to optimize the required number of sensors in a 2‐m thick surface barrier. To understand the usefulness of implementing predictive uMLR to accommodate multiple monitoring objectives, we compare several network designs, selected based on down‐sampling of existing data, with a recommended sensor design based on model simulations performed without consideration of existing data. We also added consideration of “fuzzy” design, which allows more practical guidelines for field implementation of uMLR. We found that uMLR, combined with robust decision‐making, provides a simple, flexible, and high‐quality network design for monitoring the total water stored in a surface barrier across multiple uncertain conditions.